Abstract In recent years, improvements in throughput of single cell RNA-seq have resulted in a significant increase in the number of cells profiled. The generation of single cell RNA-seq datasets comprising &gt;1 million cells is becoming increasingly common, giving rise to demands for more efficient computational workflows. Here, we present an update to our single cell RNA-seq analysis web server application, ICARUS (available at https://launch.icarus-scrnaseq.cloud.edu.au/ ) that allows effective analysis of large-scale single cell RNA-seq datasets. ICARUS v3 utilises the geometric cell sketching method to subsample cells from the overall dataset for dimensionality reduction and clustering that can be then projected to the large dataset. We then extend this functionality to select a representative subset of cells for downstream data analysis applications including differential expression analysis, gene co-expression network construction, gene regulatory network construction, trajectory analysis, cell-cell communication inference and cell cluster associations to GWAS traits. We demonstrate analysis of single cell RNA-seq datasets using ICARUS v3 of 1.3 million cells completed within the hour.

ICARUS v3, a massively scalable web server for single cell RNA-seq analysis of millions of cells